Presentation 2002/10/10
Kernel Machine Ensembles by Parallel Boosting
Michiko YAMANA, Hiroyuki NAKAHARA, Massimiliano PONTIL, Shun-ichi AMARI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose ensembles of kernel machines in which each machine is first trained on bootstrap subsets of a whole dataset and then they are combined so as to optimize objective functions. We study two different types of objective functions inspired by boosting methods, namely the exponential criterion and the maximum likelihood criterion. We also discuss the effect of additional convexity constraints to the generalization of the combination. Experimental results show that our ensemble architecture is competitive with a standard single SVM while the former reduces significantly the computational time of the training algorithm.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) ensemble learning / SVM / exponential criterion / maximum likelihood criterion
Paper # NC2002-52
Date of Issue

Conference Information
Committee NC
Conference Date 2002/10/10(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Kernel Machine Ensembles by Parallel Boosting
Sub Title (in English)
Keyword(1) ensemble learning
Keyword(2) SVM
Keyword(3) exponential criterion
Keyword(4) maximum likelihood criterion
1st Author's Name Michiko YAMANA
1st Author's Affiliation RIKEN Brain Science Institute()
2nd Author's Name Hiroyuki NAKAHARA
2nd Author's Affiliation RIKEN Brain Science Institute
3rd Author's Name Massimiliano PONTIL
3rd Author's Affiliation University of Siena
4th Author's Name Shun-ichi AMARI
4th Author's Affiliation RIKEN Brain Science Institute
Date 2002/10/10
Paper # NC2002-52
Volume (vol) vol.102
Number (no) 381
Page pp.pp.-
#Pages 5
Date of Issue